Appearance
Company Intelligence Profile — Admin Guide
This guide walks administrators through setting up and managing the Company Intelligence Profile for their organization.
Prerequisites
- Role: Owner or Admin
- Access: Admin Dashboard (sidebar → Admin Dashboard)
Initial Setup
Quick Start with Auto-Research
The fastest way to get started is to use Auto-Research, which populates your profile from publicly available information using multi-query parallel search and two-pass AI analysis:
- Navigate to Admin Dashboard → Intelligence Profile
- The setup wizard opens — on Step 1, enter your company name (domain is auto-discovered if not provided)
- Click Auto-Research
- Ghost AI runs 3 targeted search queries in parallel, scrapes discovered pages, and uses a two-pass LLM pipeline (fact extraction → profile synthesis) to suggest a description, sensitivity floor, and terms — each with a per-field confidence score
- Review and edit the suggestions, then continue through the wizard
Auto-Research is also available later from Intelligence Profile via the button in the top-right corner.
TIP
Auto-Research works best for companies with public websites. For stealth-mode or early-stage companies, skip to manual setup below.
Step 1: Open the Intelligence Profile
- Navigate to Admin Dashboard → Intelligence Profile
- If this is your first visit, the setup wizard walks you through the initial configuration
Step 2: Define Your Classification Policy
Before adding terms, establish the categories your organization needs. GhostAI ships with five built-in categories, but you can customize them or add your own.
| Built-in Category | Default Sensitivity | When to Use |
|---|---|---|
project_codename | critical | Internal project or program names |
customer_name | high | Customer and partner identifiers |
product | medium | Unreleased or internal product names |
internal_initiative | high | Strategic initiatives, M&A targets |
financial_metric | critical | Non-public financial figures |
To add a custom category:
- Click Add Category
- Enter the category name, display name, and default sensitivity
- Optionally add compliance tags (e.g.,
SOC2,ITAR,HIPAA) - Click Save
Step 3: Set the Sensitivity Floor
The Sensitivity Floor determines the minimum detection sensitivity enforced across all users.
- In the Intelligence Profile settings, locate Sensitivity Floor
- Choose a level:
- Relaxed — Only critical terms enforced
- Balanced — Medium and above enforced (recommended)
- Strict — All registered terms enforced regardless of user preference
- Click Save
TIP
Start with Balanced. You can tighten to Strict later if users are reducing their sensitivity below acceptable levels.
Step 4: Configure the Safe Terms Allowlist
Add terms that should never be flagged, even if they resemble sensitive patterns.
- Click Safe Terms in the Intelligence Profile
- Click Add Safe Term
- Enter the term and a reason (for audit trail)
- Click Save
Common safe terms include public product names, industry-standard abbreviations, and terms that frequently trigger false positives.
Managing Terms
Adding Individual Terms
- Navigate to Intelligence Profile → Sensitive Terms
- Click Add Term
- Fill in:
- Term — The word or phrase to protect
- Category — Select from your classification policy
- Sensitivity — Override the category default if needed
- Notes — Optional context for other admins
- Click Save
The term takes effect immediately for all users.
Editing Terms
- Find the term in the list (use search or category filter)
- Click the term row to expand
- Modify category, sensitivity, or metadata
- Click Save
Deleting Terms
- Find the term in the list
- Click the Delete icon
- Confirm deletion
WARNING
Deleting a term removes it from detection immediately. Messages already tokenized with this term retain their tokenization, but new messages will no longer flag the term.
Bulk Import
For organizations with large term lists, use bulk import to register many terms at once.
Via the Admin Dashboard
- Navigate to Intelligence Profile → Sensitive Terms
- Click Import
- Upload a CSV file with columns:
term,category,sensitivity(optional),notes(optional) - Review the preview — duplicates and validation errors are highlighted
- Click Import All or selectively import valid terms
CSV Format
csv
term,category,sensitivity,notes
Project Falcon,project_codename,critical,FY26 strategic initiative
Acme Corp,customer_name,high,Enterprise client
NeuralForge,product,medium,Unreleased v2 product nameVia the API
bash
curl -X POST https://api.ghostai.com/v1/intelligence-profiles/company/{orgId}/terms/bulk \
-H "Authorization: Bearer ADMIN_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"terms": [
{"term": "Project Falcon", "category": "project_codename", "sensitivity": "critical"},
{"term": "Acme Corp", "category": "customer_name", "sensitivity": "high"},
{"term": "NeuralForge", "category": "product", "sensitivity": "medium"}
]
}'The response includes counts for created, skipped (duplicates), and errored terms. See the Company Profile API Reference for full details.
Setting the Sensitivity Floor
The Sensitivity Floor is the organization's minimum detection sensitivity. Users can increase their personal sensitivity (via the Personal Intelligence Profile) but cannot reduce it below this floor.
| Floor | Effect on Users |
|---|---|
| Relaxed | Users have full control over their sensitivity dial |
| Balanced | Users who select "Relaxed" are overridden to "Balanced" |
| Strict | All users operate at maximum sensitivity regardless of preference |
When to Change the Floor
| Scenario | Recommended Floor |
|---|---|
| General-purpose organization | Balanced |
| Regulated industry (healthcare, finance, government) | Strict |
| Internal tooling with low-risk data | Relaxed |
| During a security incident or audit | Strict (temporary) |
To change the floor:
- Admin Dashboard → Intelligence Profile → Settings
- Update the Sensitivity Floor dropdown
- Click Save
The change takes effect immediately. Users whose personal sensitivity was below the new floor will see their effective sensitivity increase on their next message.
Monitoring Detection Improvement
After configuring the Company Intelligence Profile, monitor its impact through the Admin Dashboard.
Intelligence Profile Analytics
Intelligence Profile in the Admin Dashboard includes:
| Metric | Description |
|---|---|
| Total terms | Number of registered sensitive terms |
| Terms by category | Distribution across classification categories |
| Detection hits (24h) | How many times company terms were matched in the last 24 hours |
| Top matched terms | Most frequently detected terms |
| False positive rate | Percentage of company-term detections reported as false positives |
Quality Dashboard Integration
Company Intelligence Profile detections feed into the Quality Dashboard. The PII Accuracy dimension accounts for company-term false positive feedback alongside standard PII feedback.
Audit Trail
All changes to the Company Intelligence Profile are recorded in the Audit Logs:
- Term additions, modifications, and deletions
- Classification policy changes
- Sensitivity Floor changes
- Safe Terms allowlist changes
- Bulk import operations
Each audit entry includes the admin who made the change, the timestamp, and the before/after values.
Best Practices
- Start with high-value terms — Register project codenames, unreleased product names, and customer identifiers first
- Use categories consistently — Establish a classification taxonomy early and train admins to use it
- Review the Safe Terms allowlist quarterly — Terms that were safe may become sensitive over time
- Monitor false positive rate — If a term generates many false positives, consider making it more specific or adding it to the safe list
- Communicate with users — Let your team know the Company Intelligence Profile is active so they understand why new terms are being detected
- Use bulk import for onboarding — When first deploying, prepare a CSV of known sensitive terms rather than adding them one by one
Next Steps
- Company Intelligence Profile — Feature Overview
- Company Profile API Reference
- Personal Intelligence Profile — How users add their own terms
- PII Detection — Core detection pipeline
- Audit Logs — Monitoring profile changes